about your result file

[coder-yzh]

Hello, I am very interested in your work! I would like to ask about the seed setting for USPS and FASHION data sets. I want to follow your work. Is it convenient to provide the result file?
In the paper, you said that the same experimental settings were used for the balance and imbalance of the three data sets, but we could not reproduce your results. We adjusted the parameters split_merge_every_n_epochs and prior_sigma_scale, but they were still worse. Results cannot be reproduced in imbalanced versions of USPS. Could you please give some suggestions on the settings of the parameters?
Looking forward to your reply. Thank you!

[meitarronen]

Hey @coder-yzh thanks for your interest!
First of all it is important to say that we did not perform any type of seed optimization in our work.
We are checking up about the imbalanced version of USPS and will update!